AI policy

Tools may assist the work; a person is responsible for it. Every output that reaches a client, a journal or a database has been checked by the person who signs off on it. No automated result goes out unreviewed, and “the tool did it” is never an answer we give.

• Language and consistency checks: spelling, punctuation, house style and terminology. An editor accepts or rejects every suggested change.
• Reference parsing and matching: splitting references into fields and matching them to DOIs. Every match is checked against the resolved record, and anything uncertain is queried with the author.
• Metadata extraction: drafting titles, authors, affiliations and abstracts into metadata. Each field is checked against the final manuscript before deposit.
• XML conversion: converting manuscripts to JATS XML. The result is validated and reviewed against the typeset PDF.
• Similarity and image screening: flagging possible text overlap or manipulated images. A flag is a reason for a person to look, never a verdict.

• We do not generate scholarly content.
• We do not write or rewrite an author’s arguments, findings or conclusions.
• We do not produce peer review reports or editorial decisions.
• We do not create references, data, figures or citations that do not exist.

Unpublished client material is never submitted to an AI service that retains input or uses it for training. That excludes free consumer chat tools for any client material.

A tool may process unpublished material only if it runs on our own systems, or if its provider is contractually bound not to retain the input beyond the request and not to train on it. We check those terms before a tool is used, and again when they change.

On request, we tell you which tools were used on your material.

Client manuscripts, data and correspondence are never used to train or fine-tune any model, ours or anyone else’s.

On request, we give journals, publishers and databases a description of the tools used in producing a piece of work. Where a venue has its own rules on AI disclosure, we follow them and prepare whatever statement they require.

Authors remain responsible for the content they submit, including any use of AI tools in writing it, and for declaring that use where the venue requires.

When sending a manuscript to Recto, the author confirms that it is their own work and that any use of AI tools in writing it is disclosed, in the manuscript itself or in the message that accompanies it. AI tools cannot be listed as authors, and the author remains accountable for every claim, reference and figure.

If we find signs of undeclared generated content, such as invented references, we stop and tell the author or the journal. We do not fix it silently, and we do not continue until it has been resolved.

This policy follows the COPE position statement on authorship and AI tools, and the AI policies of the databases and publishers we work with. We review it when that guidance changes, and at least once a year. The date at the top shows the current version.